High data-rate infra-red optical wireless communications:implementation challenges

Transcription

1 IEEE Globecom 2010 Workshop on Optical Wireless Communications High data-rate infra-red optical wireless communications:implementation challenges Dominic O'Brien Member IEEE, Hoa Le Minh Member IEEE, Grahame Faulkner, Mike Wolf, Liane Grobe, Jianhiu Li, Olivier Bouchet Abstract Achieving high data-rates in optical wireless involves theoretical limits and practical constraints. In this paper we discuss these, with reference to two system examples. A 1.25 Gigabit/s demonstrator that has been fabricated as part of the European Community Framework 7 project OMEGA is described, and a 280Mbit/s demonstrator that is currently under development as part of the same project. In each case the compromises required, and implementation issues are discussed. a particular source within the array, and a particular detector in the receiver. Base station Switching and control Source array Tx Index Terms Optical Wireless Communications, Optical Communications Multiple element transmitter Multiple element Receiver H I. INTRODUCTION IGH data rate optical wireless systems are challenging to design and implement in the infra-red region of the optical spectrum. Transmitter power is often constrained by eye safety considerations and available receiver sensitivity is limited by the incoherent detection process. Further, the detector area that will achieve the desired bandwidth is often highly constrained by both the available components and theoretical considerations. Using multiple transmitters and multiple receivers allows higher data rates to be achieved. Links are formed between individual transmitters and receivers, and the field of view of each link is small enough to achieve sufficient power density at the receiver. There are broadly speaking two approaches to implement these systems, as shown in Figure 1. Imaging diversity[1] uses a receiver that consists of detector array with a focusing optical system, and a transmitter that uses a source array and collimating optical system. A link is formed between Dominic O'Brien and Grahame Faulkner are with the University of Oxford, Department of Engineering Science, arks Road, Oxford, OX1 3J. UK. (corresponding author Dominic O Brien: ; dominic.obrien@eng.ox.ac.uk). Hoa Le Minh is now with the School of Computing, Engineering and Information Sciences, Northumbria University Newcastle Upon Tyne NE1 8ST, United Kingdom Mike Wolf, Liane Grobe and Jianhiu Li are with the Technische Univ. Ilmenau, Germany. Olivier Bouchet is with France Telecom R&D, France. The research leading to these results has received funding from the European Community's Seventh Framework rogramme F7/ under grant agreement n also referred to as OMEGA.. This information reflects the consortiums view, and the Community is not liable for any use that may be made of any of the information contained therein. Switching and control User terminal hotodetector array (a) (b) Figure 1. (a) Angle diversity system (b) Imaging diversity system Angle diversity[2] uses an array of sources that are imaged to different angles and a receiver that uses the inverse mapping. In angle diversity separate transmitter and receiver modules are combined to give the desired coverage area. The imaging approach can create compact transceivers[3], but usually requires custom sources and detectors. Angle diversity has the theoretical advantage that rays leave and enter optical systems closer to the optical axis, and the practical advantage that there is a wider choice of components, and systems can be fabricated from commercial parts. This is balanced by the mechanical complexity and bulk of such systems. In this paper we describe aspects of the design, implementation and performance of two angle diversity systems that are being developed as part of a larger project on Home Access Networks (HANS) that is being undertaken as part of the European Community Framework 7 research programme[4]. In the next section the design constraints of angle diversity systems are discussed. Rx /10/$ IEEE 1047

2 II. ANGLE DIVERSITY SYSTEMS θ Transmitter sizes, and allowed link loss, without the effect of ambient light. In order to illustrate the constraints of system implementation and the effect on the link field of view, the next sections describe two demonstrators being developed as part of the OMEGA project. One (completed) demonstrator is designed to demonstrate Gbit/s data transfer with limited coverage, and one (underway) to demonstrate greater coverage at 280Mbit/s. on axis range h Receiver Figure 2. Single channel link Figure 2 shows a single channel optical link that forms part of the angle diversity system. A transmitter that uses a Lambertian optical source communicates with a receiver. The receiver consists of a detector area A, with an ideal optical concentrator of refractive index c det θ n which is used to increase the collection area of the receiver. The field of view of the receiver is matched to the half-angle of the Lambertian source and this link represents close to the best-case that can be achieved. It is possible to estimate the link loss given the desired field of view, detector area and concentrator gain. If the desired link loss is known attainable field of view for a single channel θ can be estimated as n 2 2 r ( n + 1) cos θ cosθ nc = Adet t r is the received power, t where 2π h sinθ, is the transmitted power, n is the Lambertian order of the transmitter beam, ( given by n = log(0.5) / log cosθ ). Link field of view -half angle(deg) ( ) detector area (mm 2 )=0.2 detector area (mm 2 )=1 detector area (mm 2 )=5 detector area (mm 2 )=10 detector area (mm 2 )= Link loss(db) Figure 3. Link loss with field of view Figure 3 shows a plot of the field of view for different detector III. GIGABIT DEMONSTRATOR As part of the OMEGA project a demonstrator that provides bi-directional Gbit ethernet transmission over a field of view of 8x25 degrees at 3m has been constructed[5]. This operates at 830nm and uses a 3 channel angle diversity approach. Details of construction and performance are shown in [5], and here we focus on the constraints and compromises in the design and implementation. A. Transmitter Any transmitter must be class 1 eye safe[6] in order that it be operated in an indoor environment. Eye safety is a function of wavelength, beam divergence and the angular subtense of the source(effectively a measure the angular extent of the image that might be formed on the human retina and therefore the power density on the retina). In addition for operation at Gbit/s rates suitable lasers and drive electronics must be available. Typical commercial circuits limit modulation currents to<100ma, which limits the optical power available from a typical GaAs laser to <30mW. Figure 4 shows the optical layout of the system. A laser emits an average power of +14dBm and this is collimated with a commercial singlet lens. The collimated beam illuminates a commercial holographic diffuser, and the size of the beam on the diffuser sets the angular subtense of the apparent source that is relevant for eyesafety. The beam divergence is set by the specifications of the holographic diffuser. Laser Singlet lens Diffuser B. Receiver Figure 4. Gigabit/s Transmitter optical layout. 1) Optoelectronics Typical receiver electronics consists of a small diameter IN or Avalanche hoto-detector (AD) connected to a commmercial transimpedance amplifier (TIA). As can be seen from the Figure 3 large detection area increases available field of view, but the capacitance associate with large detectors 1048

3 limits available bandwidth. In practice a combination of an available detector and amplifier that meets the system requirements is determined by experimental measurement, as theory does not predict available bandwidths well, and TIA circuit parameters are often unavailable. For the Gigabit system a commercial TIA and 0.2mm 2 AD integrated into a single package was used, with a. This had a measured sensitivity of ~-35dBm at 1.25Gbit/s for non-return-to-zero (NRZ) On-Off-Keyed (OOK) data with a bit-error-rate (BER) of ) Optics 10nm bandwidth interference filter Concentrating lens (12mm diameter) Sapphire half-ball lens ~3mm diameter 500 µm diameter AD Figure 5. Receiver optical system An optical system is used to increase the receiver collection area, albeit at the cost of field of view (as governed by constant radiance constraints). Achieving these theoretical limits is also challenging in terms of optical design, and the interface between optical elements and the photodetector. Figure 5 shows the optical arrangement used in this case. A glass singlet lens is combined with a sapphire ball lens to focus light onto the AD. This has a simulated optical gain (defined as the ratio of collection area to detector area) of 300 on-axis compared with a theoretical maximum of 460 for a concentrator with a refractive index of 1.5. In addition a 10nm bandwidth interference filter is used in front of the collection optics in order to reduce the ambient light noise. C. Link margin The link budget is approximately ~49dB for the Gigabit system (+14dBm transmission, -35dBm reception) and Figure 3 can be used to estimate the field of view attainable. This indicates a field of view of approximately 8 degrees (halfangle) is available for this ideal case. In practice this was reduced to 10 degrees (full-angle) to allow for additional losses, and to select a half-angle for which a holographic diffuser was available. The effect of ambient light on optical wireless systems has been extensively studied (see for instance[7]), with a wide range of different models and levels of optical noise power reported. For high speed systems the major effect is to induce additional shot noise. This is reduced by using a narrowband optical filter at the front of the receiver, although care must be taken to ensure that the passband shift of the filter with ray incidence angle does not reduce received power when the receiver is at the edge of the field of view. The effect of ambient light can be estimated, knowing the field of view and detection area. For an AD based receiver this involves (i) estimating the noise sources for a receiver of given sensitivity (ii) adding the additional shot noise from ambient light and (iii) calculating the new sensitivity with the noise sources. For the small detector areas and field of view of the receiver the ambient light noise penalty is not significant, and this was verified in measurements. System tests in bright ambient light (1200 lux as measured at the receiver), indicated the BER remained below the target D. Control and data transmission M A C L A Y E R Transmit data Received data 2-bit address 3-to-1 demux 1-to-3 splitter Clock and data recovery Clock Signal RSSI RSSI comparator and decoder Tx 1 Tx 2 Tx 3 Rx 1 Rx 2 Rx 3 Figure 6. System block diagram Figure 6 shows a block diagram of a terminal. This consists of three transmitters and three receivers. The transmitters are all on at all times, but receiver switching is required in order to switch the best received signal through to clock recovery and the system output. Maximal Ratio Combining (MRC) is the optimum way to achieve this, but requires complex circuitry to measure the channel signal to noise ratio and combine signals with the correct weight. In this case a 'select good enough' approach is taken. Each channel has a limiting amplifier to create a digital data output, and this has a threshold indicator, which is a digital line that is high if the received signal is strong enough to allow 'good' data to be received. The threshold line from each receiver channel is fed to a priority encode integrated circuit that picks the first channel that is 'good enough' and the address of this line is used to switch the received data to the CDR IC. Although not optimal this is found to be robust, and straightforward to implement. Achieving optical transmission of bursty data is challenging. Ethernet data is 8B10B line-coded to obtain a DC balanced data stream so baseline wander can be controlled relatively easily. However, when a new Ethernet packet is sent the system must respond quickly enough to transmit the data, receive it, threshold and limit the received signal, select the desired receiver of the three angle diversity channels, route 1049

4 this signal to clock recovery, and finally recover the clock and data. Within the Ethernet packet there is a preamble, which contains a 'K' code at the end which signifies the beginning of data and the preamble must be long enough to ensure that the system produces valid data by the start if this code. This requires careful design of the AC coupling between stages, and careful component selection to ensure a fast enough response. In the Gigabit system the receiver system 'wakes-up' in approximately 500ns or so, corresponding to a preamble of 62 bytes or so. This requires a custom length preamble, and the Gigabit system the Ethernet interface was designed to continually transmit an 'idle' signal to circumvent this problem. In the 280Mbit/s system a custom length preamble is will be added to the Ethernet packet to ensure the system will respond correctly to burst data. For the systems described here each terminal has an identical downlink and uplink, which allow simultaneous transmission on both links, and the MAC layer [8] implements a Time Division Multiple Access (TDMA) that can support this. For the second system half-duplex transmission will be implemented in the MAC layer. In the longer term these angle diversity systems have the capability to support Space Division Multiple Access (SDMA) offering the potential for extremely high throughputs, but this requires much more complex control than was implemented in these demonstrators. 1) System performance IV. SYSTEM II The system described previously has high data rates but limited field of view. A second seven channel system, designed to achieve a 90 degree field of view for a base station 3m above a receiver plane is currently under development. In this case the target line rate is approximately 280Mbit/s, the reduction in data rate allowing increased transmission power and detector area. This system is briefly described in the following sections, and the changes in design discussed. A. Transmitter A reduction of data rate allows higher current commercial drivers, and lasers with higher emission power to be obtained, compared to the Gigabit system. In this case approximately 80mW of power is available for an 860nm laser operating at ~300Mbit/s NRZ OOK. The optical layout used is similar to that shown in Figure 4. B. Receiver The major benefit in reducing receiver bandwidth is that larger area detectors are available, and the range of detectors available increases substantially. To the first order bandwidth might be modelled as inversely proportional to detector area due to the area dependent capacitance, and sensitivity should also increase in the same manner. Using these arguments an area of 0.8mm 2 and a sensitivity of~-41dbm might be achieved using scaling of the Gigabit system results. In practice a 3mm 2 AD in combination with a commercial TIA with a sensitivity of ~-38dBm at ~300Mbit/s (NRZ-OOK with the required BER) was chosen. The lower sensitivity than expected is more than compensated by the increased area relative to expectation. In order to simplify receiver optical design and to increase the collection area of the optical system a custom optical concentrator was specified for the receiver. Figure 8 shows a cross section of a receiver channel, including the concentrator. The concentrator has a simulated gain of ~32, compared with a maximum theoretical gain of~40, and this remains relatively constant with incident angle, whereas the optical system in Figure 5 shows a ~50% fall of in gain with angle. This, and the ease of integration are the main advantages of this approach. Concentrator Figure 7. User terminal Figure 7 shows a picture of a single user terminal. The three transmitter and three receiver channels can be clearly be seen. The link was tested and operated up to a range of 4m on axis, maintaining the required BER. Off-axis tests were made at a range of 3m, and a field of view of 8x25 degrees was achieved. More complete details of performance and measurements can be found in [5]. Avalanche hotodetector Figure 8. Single channel receiver schematic An interference filter is used to reduce the effect of ambient light in this system also. The passband shift for signals that enter the receiver at the edge of the angular field of view must be accounted for and a filter width of 23nm was chosen to ensure high transmission of the signal for all entry angles. 1050

5 C. Link budget and ambient light The expected system link margin is 57dB which for a 3mm 2 AD corresponds to an approximately 25 degree half-angle field of view being feasible (see Figure 3). In practice a field of view of 15 degrees half-angle was chosen. For this system the increased detector area and field of view mean that effect of ambient light must be considered, although it is difficult to accurately estimate. The effect of the shot noise due to ambient light was modelled as outlined in section III.C. Figure 9 shows the field of view vs receiver sensitivity (with and without ambient light) for a transmitter power of 80mW. These show that with a (high) level of ambient light of 40mW/Sr/m 2 /nm the shot noise from this light begins to dominate and there is little benefit from an intrinsically more sensitive receiver. This effect is increased by the need to widen the bandwidth of the interference filter for larger field of view values. A field of view of 15 degrees half-angle, which is chosen for the demonstrator. It should be noted that this modelling depends on parameters that are not easy to determine and experimental measurement is required to verify this, but the general trend shown on these graphs is as would be expected. Link field of view -half angle(deg) Receiver sensitivity (dbm) -35 No Ambient light Ambient light Figure 9. Effect of ambient light on field of view. The receiver sensitivity axis refers to the receiver sensitivity 'in the dark'. -40 collection, decreased concentrator gain, and the need to increase optical filter bandwidth. For the systems considered here 15 degrees half-angle appears to be approximately the limit of what is easily achievable. This work shows that angle diversity systems can achieve high bandwidth, but that this is at the cost of complex systems. Imaging diversity allows more compact transceivers and will be required for systems beyond 1Gbit/s, as the number of separate channels to achieve coverage required increases. Such receivers are being developed[9], showing that data rates in the range of 1-10Gbits/s are feasible. Such high data rate, low latency, systems have the potential to provide ultra-high data rate 'hotspot' connectivity in future wireless systems. REFERENCES [1]G. Yun and M. Kavehrad, "Spot-diffusing and fly-eye receivers for indoor infrared wireless communications," in Wireless Communications, Conference roceedings., 1992 IEEE International Conference on Selected Topics in, 1992, pp [2]J. B. Carruthers and J. M. Kahn, "Angle diversity for nondirected wireless infrared communication," IEEE Transactions on Communications, vol. 48, pp , [3]D. C. O'Brien, G. E. Faulkner, E. B. Zyambo, K. Jim, D. J. Edwards,. Stavrinou, G. arry, J. Bellon, M. J. Sibley, V. A. Lalithambika, V. M. Joyner, R. J. Samsudin, D. M. Holburn, and R. J. Mears, "Integrated transceivers for optical wireless communications," IEEE Journal of Selected Topics in Quantum Electronics, vol. 11, pp , Jan.-Feb [4]J.. Javaudin, M. Bellec, D. Varoutas, and V. Suraci, "OMEGA ICT project: Towards convergent Gigabit home networks," in ersonal, Indoor and Mobile Radio Communications, IMRC IEEE 19th International Symposium on, 2008, pp [5]H. Le-Minh, O'Brien-Dc, G. Faulkner, M. Wolf, L. Grobe, J. Lui, and O. Bouchet, "A 1.25Gbit/s Indoor optical wireless demonstrator," Submitted to hotonics Technology Letters, [6]IEC Safety of laser products part 1: British Standards Institution, [7]A. C. Boucouvalas, "Indoor ambient light noise and its effect on wireless optical links," Iee roceedings-optoelectronics, vol. 143, pp , [8]. orcon, "Optical Wireless MAC Specification," in [9]J. Zeng, V. Joyner, J. Liao, S. L. Deng, and Z. R. Huang, "A 5Gb/s 7- Channel Current-mode Imaging Receiver Front-end for Free-Space Optical MIMO," nd Ieee International Midwest Symposium on Circuits and Systems, Vols 1 and 2, pp , V. DISCUSSION AND CONCLUSIONS In this paper we report demonstrations at 1.25Gbit/s and designs to operate at 280Mbit/s. Angle diversity systems are used due to constraints in available components, and there are many design compromises that are made due to this constraint. There are two broad conclusions from this work; Reducing bandwidth brings greater than expected benefits in terms of increased detection area and hence link margin, as well as a greater choice of components and ease of implementation. Using more complex modulation schemes to achieve high data rates may therefore have benefits as low bandwidth optoelectronic components can be used, with the associated increase in link margin that is seen in these examples. Further analysis of how the link margin scales with bandwidth is required to determine if this is the case. Increasing the field of view of a single channel involves multiple penalties in terms of increased ambient light 1051